Predictors of Return to Work After Concussion Injury; Results From a Multiple Logistic Regression Model of 1, 103 Patients Over 3 Years in a Single Canadian Institution. (16th November 2020)

• Record Type: Journal Article
• Authors: Wilson, Jamie R.F; Badhiwala, Jetan H; Martin, Allan R; Moghaddamjou, Ali; Massicotte, Eric M
• Is Part Of: Neurosurgery. Volume 67(2010)Supplement 1
• Publication Date: 2020-11-16
• Subjects: Nervous system -- Surgery -- Periodicals; 617.48005
• DOI: 10.1093/neuros/nyaa447_482
• Languages: English
• ISSNs: 0148-396X

Abstract

Abstract: INTRODUCTION: Identifying clinical indicators that significantly predict the likelihood of return to work after concussion has proved difficult, with a number of discrepancies in the literature. METHODS: All patients who underwent outpatient treatment for concussion injury during the years 2017–2019 at a single Canadian institution were included. Simple demographics [age, previous concussion injuries, previous mental illness, work status, marriage status, educational level, primary diagnosis, time to treatment, treatment duration] were collected with initial scores for PHQ9, Rivermead, GAD7 and MOCA assessments. Categorical and numerical variables were used to create a multiple logistic regression model with variables were removed if estimable, >0.10 significance value or if multivariable fractional polynomial analysis did not detect interaction. RESULTS: 1103 patients were identified, with a mean age 43.9 [43.1-44.7]. 62% were female, and mean duration of treatment was 2.8[2.7-2.9] months. 599 [54%] reported a return to any form of work after treatment. Logistic regression modelling revealed part-time (OR 4.6[1.3-16.1]; P = . 018) or full-time (OR 4.4[1.5-12.8]; P = . 008) work at the time of injury, or female sex (OR 1.8[1.1-3.1]; P = . 031) were the strongest predictors of return to work. Treatment duration (OR 1.5 [1.3-1.8] per unit month increase; P < . 001) was positively correlated with return to work with the total Rivermead score demonstrating negative correlation (OR 0.96 [0.93-0.99] per unit score increase; P = . 007). After regression analysis age, time to treatment, marriage status, previous psychological illness, a history of previous concussion, initial MOCA, PHQ9 and GAD7 scores were not significant predictors. CONCLUSION: Working at the time of injury, female sex, longer treatment duration and lower Rivermead score appear to be significant predictors of return to work after concussion injury. Further work could include machine-learning models to improve the acuity of the model.